microarray_normalization_Help!!!

Jose de las Heras josenet at tiscali.co.uk
Sat Apr 22 05:52:32 EST 2006


You could attempt to normalise with Excel. But it's not the best way.
If you're going to analyse microarrays, I recommend you use something like 
Limma (linear models for microarray data).

You can use Limma to take your raw data and so several type of diagnostic 
tests to check the quality. Then you can apply a number of types of 
background correction, normalisation (both within each array for Cy3/Cy5, 
and between arrays), and producing a list of differentially expressed genes 
with stats etc.
And all kinds of plots, highlighting genes individually or in groups...
In addition, the sorting and subsetting abilities of limma are more powerful 
and faster than Excel.

Ok, the downside is that limma is command-line based. So you do have to 
spend a little time learning how to use it. But it's easy. There's a users 
guide that takes you step by step through several worked examples showing 
you how to do most things you'd want to do, and in a couple of days you'd be 
running with your data. It'll be a good investment.

Limma is a package from the BioConductor project. BioConductor is a group of 
packages designed for the analysis of microarray data. You'll find apart 
from limma other tools that will allow you to do clustering analysis, 
linking to gene ontology databases and all sorts of stuff. I am only really 
familiar with Limma.
All these are based around the statistical-oriented programming language R.

All these are free, with extensive help documentation and there's also a 
BioConductor forum where you can get very useful help if you get stuck 
somewhere.

All you need to do is go to:

http://cran.r-project.org/

and download and install the latest version of R for your platform (windows, 
mac, unix)

then, from the same page above, on the left you find a number of links. One 
is "packages". Go there, and download the zip file for "limma". Next, run R, 
and from the top menu select "install package from zip file", and select the 
limma one. You're done. Then check the user's guide included in the limma 
folder and start working through the example.

It's also useful to go to the BioConductor site:

http://www.bioconductor.org/

The BioConductor site has lots of information and there you can find the 
link to the BioC forum I mentioned. It gets updated less frequently than the 
info in the R project site above, so it's best to get your R and Limma from 
the first website.

That would be my preferred option, and one that will serve you well forever.

If you find the command-line version of Limma a bit hard going, there's a 
version with a graphical interface (GUI) called limmaGUI. You can get it 
from:

http://bioinf.wehi.edu.au/limmaGUI/

If you use windows, you can download a single file that will install R 
version 2.1.1 and LimmaGUI and all the packages to make it work together in 
one go.
This is the simplest way to get started, in 20 minutes, you'll be up and 
running with your data normalised etc. The problem I see is that the options 
are limited, compared to straight command-line based limma. But you can get 
around it by typing your own commands ina window that you can open from 
LimmaGUI. Still... if you're going to use limma commands I'd rather do it 
all from the beginning, but... you may prefer it, check it out.

In addition to that, I find the TM4 suite for microarray analysis from TIGR 
very useful.And it's also free. Check it out at:

http://www.tm4.org/

There you get SpotFinder, which you can sue to quantitate your images (you 
won't need that as I guess you use GenePix... I also use GenePix now, but 
started with SpotFinder, and I still go back to SpotFinder a lot. I like how 
you can click on spots on a plot and it'll show you the actual spot 
intensities, annotation etc... I know GenePix does something similar, but I 
like SpotFinder's evrsion better).

You also get MIDAS. MIDAS allows you to normalise data and so some filtering 
based on a number of conditions. MIDAS takes the output from SpotFinder, but 
you can convert your GPR files to MEV format (the one used by SpotFinder) 
using their tool ExpressConverter, and then use that for MIDAS. Apparently 
the new version of MIDAS (notout yet) will take GPR files directly, and 
other nice improvements, but they haven't told me when it'll be out.

And you also get TMEV. You can use MEV and GPR files as input. TMEV does 
clustering analysis and it's quite nice.

I mainly use Limma to start, and later use TMEV (either from GPR files of 
the MEV ones) if I want to do clustering etc.

I am very slowly writing a little "easy" guide to use these programs to d 
some simple data normalisation and analysis, for use in our lab... it saves 
me a lot of time if I can give it to a new person and they familiarise 
themselves before we start. It's still unfinished and has many gaps.. but 
the Limma and LimmaGUI part is pretty much complete.. if you want it I'll 
email it to you.

I hope this helps!

good luck with you arrays

Jose





"gberna" <gberna at gmail.com> wrote in message 
news:1145462502.719792.29860 at u72g2000cwu.googlegroups.com...
> Dears,
> I have some problems about how to analyze my data:
> I'm processing some microarray with protein .
> On  every slide, I made an hybridization on slide with peptides and my
> antibody  was  colored  with Cy3 and Cy5.
> In this case the the spots would have to be the same one (becouse cy3
> and cy5 are the same one), in fact I see a yellow spot
> Query:
> How  can I process this data?
> Using excel I calculated log 2 (cy3median-bkg/cy5median-bkg)
> How can I normalize the data?
> Can you help me?
>
> I want to see the report between for example the prtA_phosfo/prtA (2
> peptides on slides)and I don't know which data I must consider.
>
> I hope that someone have understood this delirious post.
> thanks,
>
> guido
>
> PS
> sorry for my english
> I use GPR file
> 




More information about the Methods mailing list